MODIFICATION OF SWITCH PROBABILITY IN FLOWER POLLINATION ALGORITHM THROUGH STATISTICAL INFERENCE : ANALYSIS OF VARIANCE

• Main Author: Hanif Sulaiman, Fadhil
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/73124
• Institution: Institut Teknologi Bandung

Flower Pollination Algorithm or commonly abbreviated as FPA is an algorithm whose functions is to solve optimization problems. In this algorithm the search for optimal value is carried out with two categories, namely global and local pollination. The probability of occurrence of the two categories is controlled by a parameter known as the switch probability. The value selection of the switch probability will affect the results of the algorithm, namely the number of iterations needed to achieve the optimum value. In general, the selected switch probability value is 0.8. In this final project, FPA will be modified with various switch probability values, namely constant values (0.6 - 0.9), following the Double Exponential rule, and following the distribution of Beta(18.2). In carrying out this study, FPA will be simulated with various rules. Then will be analyzed with descriptive statistics and inference. The purpose of this research is to test whether there is a significant difference between the switch probability treatments used. Then it will be search for which treatment produces the fewest iterations. The methods used in this final project are literature studies, computer-assisted simulations, and analysis with statistical processing. In this study, there were several results that were quite different from the results in the reference. Therefore, the result is carried out based on some achievement of minimum success rate. The analysis steps carried out are erasing outliner; descriptive statistics both analytically and graphically; and statistical inference namely ANOVA. To meet the requirements of ANOVA, the Shapiro Wilk normality test and the Bartlett variance convenience test were performed. Tukey ladder transformation was also performed to obtain more symmetrical data. After the ANOVA conducted, the Tukey test was also carried out to detect pair data that had different means. Then to get the appropriate switch probability, a rating is given to the sample average. From this study, the results show that the switch probability 0.9 and Beta(18.2) gives a minimum iteration at a success rate of 100%. Whereas for other criteria or objective functions with large dimensions, the switch probability 0.6 gives the minimum iteration. Keywords : FPA, switc